Biometry Research Group, Division of Cancer Prevention, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892-8574, USA.
Clin Trials. 2010 Feb;7(1):69-77. doi: 10.1177/1740774509356461.
Often in randomized controlled trials of cancer screening, cause of death is determined by a mortality review committee. However, little is known regarding how findings from mortality review compare to those from death certificates alone.
To examine the results of four different U. S. trials of cancer screening when death certificate data only were used, as compared to results using all available mortality review information.
Trials included were the Health Insurance Plan of New York breast screening trial (HIP), the Minnesota trial of fecal occult blood testing, and the Johns Hopkins and Mayo Lung Projects, which each examined chest x-ray and sputum cytology. The sensitivity, specificity, positive and negative predictive values, and Cohen's kappa for death certificates were calculated for all arms of all trials. Separate intention-to-screen analyses were conducted for each trial using cause of death information from either death certificates alone or full mortality review data.
Generally there was excellent agreement between the death certificates and the mortality review committee as to the underlying cause of death (kappa >0.85 in all cases); death certificate agreement was similar between arms in all trials. Modest changes in the screening effectiveness estimates were observed when mortality review information was utilized, ranging from a 9% decrease to a 2% increase in the calculated mortality rate ratios. However, in one instance (HIP) a statistically significant benefit of screening was observed when mortality review committee data were used (rate ratio (RR) 0.77, 95% confidence interval (CI) 0.62- 0.95) but not when death certificate data were used (RR 0.82, 95% CI 0.65-1.03).
Although considered to be the gold standard, even carefully conducted mortality review may result in errors in cause of death assignment.
For each trial, results were similar regardless of the source of cause of death information.
在癌症筛查的随机对照试验中，死因通常由死亡审查委员会确定。然而，对于死亡率审查的结果与单独使用死亡证明的结果相比，人们知之甚少。
当仅使用死亡证明数据时，检查四项不同的美国癌症筛查试验的结果，与使用所有可用死亡率审查信息时的结果进行比较。
纳入的试验包括纽约健康保险计划的乳腺筛查试验（HIP）、明尼苏达州粪便潜血检测试验以及约翰霍普金斯和梅奥肺部项目，每个试验都检查了胸部 X 光和痰液细胞学。为所有试验的所有臂计算了死亡证明的敏感性、特异性、阳性和阴性预测值以及 Cohen's kappa。对于每个试验，使用单独的死亡证明或完整的死亡率审查数据的死因信息，分别进行了意向筛查分析。
在所有情况下，死亡证明与死亡率审查委员会对死亡根本原因的判断之间通常具有极好的一致性（kappa >0.85）；在所有试验中，死亡证明在各臂之间的一致性相似。当使用死亡率审查信息时，观察到筛查效果估计值的适度变化，从计算的死亡率比降低 9%到增加 2%。然而，在一个实例（HIP）中，当使用死亡率审查委员会的数据时，观察到筛查具有统计学上的显著益处（RR 0.77，95%置信区间（CI）0.62-0.95），而当使用死亡证明数据时则没有（RR 0.82，95% CI 0.65-1.03）。
即使是精心进行的死亡率审查，也可能导致死因分配错误，被认为是金标准。
对于每个试验，结果相似，无论死因信息的来源如何。
